MEG Without a Magnetically Shielded Room: Progress and Remaining Challenges

Magnetoencephalography (MEG) requires a magnetically shielded room (MSR) in every current clinical deployment. The MSR is a multi-layer construction of mu-metal and aluminium panels forming a room-within-a-room, typically costing 500,000 to 1.5 million Australian dollars to install, requiring dedicated infrastructure, and restricting MEG to specialist tertiary centres. The question of whether quantum sensors might eventually eliminate this requirement generates genuine interest. This post addresses what the physics and current technology actually allow, and what remains out of reach.

Why an MSR is currently needed

The neural magnetic signals measured in MEG are in the range of 10 to 1000 femtotesla. Urban electromagnetic noise from building wiring, HVAC equipment, passing vehicles, and elevator motors is typically in the range of 1 to 1000 nanotesla at 1 to 100 Hz, which is six to nine orders of magnitude larger than the neural signal. The MSR provides passive attenuation of approximately 80 to 100 dB at 1 Hz (the relevant clinical MEG frequency range), reducing the external noise entering the measurement space to the picotelsa to femtotesla range. Even after this attenuation, the residual noise inside the MSR often exceeds the neural signal amplitude, which is why MEG traditionally used SQUID sensors with femtotesla per root hertz sensitivity to detect signals in this noise environment.

The MSR does not eliminate interference; it reduces it to a level where sensitive enough sensors can still detect the neural signal. Removing the MSR entirely would require either sensors with much better sensitivity than current room-temperature options, or active interference rejection capable of achieving 80 to 100 dB attenuation against broadband urban noise, or both.

Active shielding: what it achieves today

Active noise cancellation for MEG uses reference magnetometers placed away from the subject but within the same room, to measure the ambient interference field and generate a correction signal through cancellation coils. This approach has been demonstrated to achieve 40 to 60 dB of broadband rejection from 1 to 100 Hz in controlled settings using dense reference sensor arrays and adaptive filtering algorithms. Adding active cancellation to a low-attenuation passive enclosure (50 dB passive + 40 dB active) can in principle approach the total shielding provided by a conventional MSR.

In practice, active cancellation performance degrades in several important ways. The interference field must be spatially correlated across the scale of the subject's head (approximately 20 centimetres) for the reference channels to predict the interference at the MEG sensor positions accurately. In environments with interference sources at varying distances and directions (building wiring at multiple floors, moving vehicles on a nearby road), the spatial structure of the interference at the sensor array changes dynamically in ways that reference channel correction cannot fully track. Demonstrated rejection in real urban clinical environments is typically 20 to 35 dB rather than the 40 to 60 dB achieved in controlled conditions.

Where room-temperature quantum sensors fit

Current NV-center and OPM sensors for MEG achieve approximately 10 to 50 femtotesla per root hertz sensitivity in low-noise (MSR) environments, which is sufficient for the strongest MEG signals (auditory evoked fields, motor-related signals) but marginal for the weakest (single-neuron oscillatory activity, some epileptiform events). In an unshielded environment with 20 to 35 dB active rejection, the effective noise floor seen by a 20 fT per root hertz sensor is still 200 to 600 fT per root hertz, which is above the amplitude of most diagnostically relevant neural sources.

This gap between active rejection performance and the requirement to detect weak neural signals in an unshielded clinical environment is real and is not closed by current technology. Incremental improvements in active cancellation algorithm performance and sensor sensitivity each move the boundary, but the gap between where those improvements bring the system and where it needs to be for broadband MEG without an MSR is measured in tens of decibels, not one or two.

The limited but realistic near-term case

There is a narrower application where room-temperature sensors with active shielding can provide useful MEG-adjacent measurements in non-MSR environments: detection of strong focal magnetic signals from implanted neural devices, high-amplitude pathological activity (large interictal spikes with field amplitudes above 500 fT), or research studies where signal averaging over many thousands of trials is feasible and the signal of interest is synchronised to a known stimulus. These are not the full MEG diagnostic application, but they represent real clinical value in specific contexts.

For a well-managed suburban or rural clinical site (not a metropolitan hospital), ambient noise levels can be 10 to 20 dB lower than urban centres, reducing the active cancellation requirement. At 50 to 60 dB of combined passive-plus-active attenuation, the noise floor seen by a 20 fT per root hertz sensor falls to 60 to 200 fT per root hertz, which is compatible with averaged evoked field measurements for auditory, somatosensory, and motor tasks. This is the application context that DeteQt's sensor array work is directed at for medical imaging, not replacing full clinical MSR-based MEG installations.

Honest assessment of the pathway to unshielded MEG

Achieving broadband clinical MEG (the full diagnostic capability) without an MSR in a typical hospital environment would require: active rejection improving to 60 to 80 dB across 1 to 100 Hz in dynamic interference conditions; sensor sensitivity improving to 5 to 10 fT per root hertz in ambient fields; and demonstration of stable operation across the daily variation in hospital interference environment. None of these individually appears near-term impossible, but achieving all three simultaneously in a reliable clinical instrument has not been demonstrated.

We think the realistic trajectory is a reduction in MSR cost and complexity rather than elimination. A lighter-weight single-layer mu-metal enclosure providing 50 to 60 dB of passive attenuation (versus 80 to 100 dB for a full MSR), combined with active cancellation and room-temperature sensors, would reduce installation cost by perhaps a factor of five to eight compared to a conventional MSR while delivering adequate performance for a subset of clinical applications. Whether that subset covers the primary diagnostic use cases (pre-surgical epilepsy localisation, tumour mapping, disorder of consciousness assessment) is an open clinical research question, not one we can answer from sensor specifications alone.

We are not building a complete MEG system. DeteQt's contribution to this space is the sensor array component, specifically compact NV-center magnetometer arrays that can be positioned conformally around the scalp and operated at room temperature. The integration, clinical validation, regulatory pathway, and health economics questions are for system developers and clinical research groups to address.